Gita Sukthankar

Shengnan Hu, Ce Zheng, Zixiang Zhou et al.

Human-centric visual understanding is an important desideratum for effective human-robot interaction. In order to navigate crowded public places, social robots must be able to interpret the activity of the surrounding humans. This paper addresses one key aspect of human-centric visual understanding, multi-person pose estimation. Achieving good performance on multi-person pose estimation in crowded scenes is difficult due to the challenges of occluded joints and instance separation. In order to tackle these challenges and overcome the limitations of image features in representing invisible body parts, we propose a novel prompt-based pose inference strategy called LAMP (Language Assisted Multi-person Pose estimation). By utilizing the text representations generated by a well-trained language model (CLIP), LAMP can facilitate the understanding of poses on the instance and joint levels, and learn more robust visual representations that are less susceptible to occlusion. This paper demonstrates that language-supervised training boosts the performance of single-stage multi-person pose estimation, and both instance-level and joint-level prompts are valuable for training. The code is available at https://github.com/shengnanh20/LAMP.

Muhammad Junaid Khan, Syed Hammad Ahmed, Gita Sukthankar

The StarCraft II Multi-Agent Challenge (SMAC) was created to be a challenging benchmark problem for cooperative multi-agent reinforcement learning (MARL). SMAC focuses exclusively on the problem of StarCraft micromanagement and assumes that each unit is controlled individually by a learning agent that acts independently and only possesses local information; centralized training is assumed to occur with decentralized execution (CTDE). To perform well in SMAC, MARL algorithms must handle the dual problems of multi-agent credit assignment and joint action evaluation. This paper introduces a new architecture TransMix, a transformer-based joint action-value mixing network which we show to be efficient and scalable as compared to the other state-of-the-art cooperative MARL solutions. TransMix leverages the ability of transformers to learn a richer mixing function for combining the agents' individual value functions. It achieves comparable performance to previous work on easy SMAC scenarios and outperforms other techniques on hard scenarios, as well as scenarios that are corrupted with Gaussian noise to simulate fog of war.

Amirarsalan Rajabi, Mehdi Yazdani-Jahromi, Ozlem Ozmen Garibay et al.

With the recent growth in computer vision applications, the question of how fair and unbiased they are has yet to be explored. There is abundant evidence that the bias present in training data is reflected in the models, or even amplified. Many previous methods for image dataset de-biasing, including models based on augmenting datasets, are computationally expensive to implement. In this study, we present a fast and effective model to de-bias an image dataset through reconstruction and minimizing the statistical dependence between intended variables. Our architecture includes a U-net to reconstruct images, combined with a pre-trained classifier which penalizes the statistical dependence between target attribute and the protected attribute. We evaluate our proposed model on CelebA dataset, compare the results with a state-of-the-art de-biasing method, and show that the model achieves a promising fairness-accuracy combination.

Shengnan Hu, Gita Sukthankar

This paper presents a new approach for predicting team performance from the behavioral traces of a set of agents. This spatiotemporal forecasting problem is very relevant to sports analytics challenges such as coaching and opponent modeling. We demonstrate that our proposed model, Spatial Temporal Graph Convolutional Networks (ST-GCN), outperforms other classification techniques at predicting game score from a short segment of player movement and game features. Our proposed architecture uses a graph convolutional network to capture the spatial relationships between team members and Gated Recurrent Units to analyze dynamic motion information. An ablative evaluation was performed to demonstrate the contributions of different aspects of our architecture.

Ayesha Enayet, Gita Sukthankar

Conflict prediction in communication is integral to the design of virtual agents that support successful teamwork by providing timely assistance. The aim of our research is to analyze discourse to predict collaboration success. Unfortunately, resource scarcity is a problem that teamwork researchers commonly face since it is hard to gather a large number of training examples. To alleviate this problem, this paper introduces a multi-feature embedding (MFeEmb) that improves the generalizability of conflict prediction models trained on dialogue sequences. MFeEmb leverages textual, structural, and semantic information from the dialogues by incorporating lexical, dialogue acts, and sentiment features. The use of dialogue acts and sentiment features reduces performance loss from natural distribution shifts caused mainly by changes in vocabulary. This paper demonstrates the performance of MFeEmb on domain adaptation problems in which the model is trained on discourse from one task domain and applied to predict team performance in a different domain. The generalizability of MFeEmb is quantified using the similarity measure proposed by Bontonou et al. (2021). Our results show that MFeEmb serves as an excellent domain-agnostic representation for meta-pretraining a few-shot model on collaborative multiparty dialogues.

Ayesha Enayet, Gita Sukthankar

Social coding platforms, such as GitHub, serve as laboratories for studying collaborative problem solving in open source software development; a key feature is their ability to support issue reporting which is used by teams to discuss tasks and ideas. Analyzing the dialogue between team members, as expressed in issue comments, can yield important insights about the performance of virtual teams. This paper presents a transfer learning approach for performing dialogue act classification on issue comments. Since no large labeled corpus of GitHub issue comments exists, employing transfer learning enables us to leverage standard dialogue act datasets in combination with our own GitHub comment dataset. We compare the performance of several word and sentence level encoding models including Global Vectors for Word Representations (GloVe), Universal Sentence Encoder (USE), and Bidirectional Encoder Representations from Transformers (BERT). Being able to map the issue comments to dialogue acts is a useful stepping stone towards understanding cognitive team processes.

Muhammad Junaid Khan, Gita Sukthankar

This paper introduces SC-Phi2, a fine-tuned StarCraft II small language model for macromanagement tasks. Small language models, like Phi2, Gemma, and DistilBERT, are streamlined versions of large language models (LLMs) with fewer parameters that require less power and memory to run. To teach Microsoft's Phi2 model about StarCraft, we create a new SC2 text dataset with information about StarCraft races, roles, and actions and use it to fine-tune Phi-2 with self-supervised learning. We pair this language model with a Vision Transformer (ViT) from the pre-trained BLIP-2 (Bootstrapping Language Image Pre-training) model, fine-tuning it on the MSC replay dataset. This enables us to construct dynamic prompts that include visual game state information. Unlike the large models used in StarCraft LLMs such as GPT-3.5, Phi2 is trained primarily on textbook data and contains little inherent knowledge of StarCraft II beyond what is provided by our training process. By using LoRA (Low-rank Adaptation) and quantization, our model can be trained on a single GPU. We demonstrate that our model performs well at micromanagement tasks such as build order and global state prediction with a small number of parameters.

Syed Hammad Ahmed, Shengnan Hu, Gita Sukthankar

Natural language supervision has been shown to be effective for zero-shot learning in many computer vision tasks, such as object detection and activity recognition. However, generating informative prompts can be challenging for more subtle tasks, such as video content moderation. This can be difficult, as there are many reasons why a video might be inappropriate, beyond violence and obscenity. For example, scammers may attempt to create junk content that is similar to popular educational videos but with no meaningful information. This paper evaluates the performance of several CLIP variations for content moderation of children's cartoons in both the supervised and zero-shot setting. We show that our proposed model (Vanilla CLIP with Projection Layer) outperforms previous work conducted on the Malicious or Benign (MOB) benchmark for video content moderation. This paper presents an in depth analysis of how context-specific language prompts affect content moderation performance. Our results indicate that it is important to include more context in content moderation prompts, particularly for cartoon videos as they are not well represented in the CLIP training data.

Jack Vice, Natalie Ruiz-Sanchez, Pamela K. Douglas et al.

In humans, intrinsic motivation is an important mechanism for open-ended cognitive development; in robots, it has been shown to be valuable for exploration. An important aspect of human cognitive development is $\textit{episodic memory}$ which enables both the recollection of events from the past and the projection of subjective future. This paper explores the use of visual episodic memory as a source of intrinsic motivation for robotic exploration problems. Using a convolutional recurrent neural network autoencoder, the agent learns an efficient representation for spatiotemporal features such that accurate sequence prediction can only happen once spatiotemporal features have been learned. Structural similarity between ground truth and autoencoder generated images is used as an intrinsic motivation signal to guide exploration. Our proposed episodic memory model also implicitly accounts for the agent's actions, motivating the robot to seek new interactive experiences rather than just areas that are visually dissimilar. When guiding robotic exploration, our proposed method outperforms the Curiosity-driven Variational Autoencoder (CVAE) at finding dynamic anomalies.

Muhammad Junaid Khan, Syed Hammad Ahmed, Gita Sukthankar

We present a novel method aimed at enhancing the sample efficiency of ensemble Q learning. Our proposed approach integrates multi-head self-attention into the ensembled Q networks while bootstrapping the state-action pairs ingested by the ensemble. This not only results in performance improvements over the original REDQ (Chen et al. 2021) and its variant DroQ (Hi-raoka et al. 2022), thereby enhancing Q predictions, but also effectively reduces both the average normalized bias and standard deviation of normalized bias within Q-function ensembles. Importantly, our method also performs well even in scenarios with a low update-to-data (UTD) ratio. Notably, the implementation of our proposed method is straightforward, requiring minimal modifications to the base model.

Syed Hammad Ahmed, Muhammad Junaid Khan, Gita Sukthankar

Due to the rise in video content creation targeted towards children, there is a need for robust content moderation schemes for video hosting platforms. A video that is visually benign may include audio content that is inappropriate for young children while being impossible to detect with a unimodal content moderation system. Popular video hosting platforms for children such as YouTube Kids still publish videos which contain audio content that is not conducive to a child's healthy behavioral and physical development. A robust classification of malicious videos requires audio representations in addition to video features. However, recent content moderation approaches rarely employ multimodal architectures that explicitly consider non-speech audio cues. To address this, we present an efficient adaptation of CLIP (Contrastive Language-Image Pre-training) that can leverage contextual audio cues for enhanced content moderation. We incorporate 1) the audio modality and 2) prompt learning, while keeping the backbone modules of each modality frozen. We conduct our experiments on a multimodal version of the MOB (Malicious or Benign) dataset in supervised and few-shot settings.

Syed Hammad Ahmed, Muhammad Junaid Khan, H. M. Umer Qaisar et al.

Online video platforms receive hundreds of hours of uploads every minute, making manual content moderation impossible. Unfortunately, the most vulnerable consumers of malicious video content are children from ages 1-5 whose attention is easily captured by bursts of color and sound. Scammers attempting to monetize their content may craft malicious children's videos that are superficially similar to educational videos, but include scary and disgusting characters, violent motions, loud music, and disturbing noises. Prominent video hosting platforms like YouTube have taken measures to mitigate malicious content on their platform, but these videos often go undetected by current content moderation tools that are focused on removing pornographic or copyrighted content. This paper introduces our toolkit Malicious or Benign for promoting research on automated content moderation of children's videos. We present 1) a customizable annotation tool for videos, 2) a new dataset with difficult to detect test cases of malicious content and 3) a benchmark suite of state-of-the-art video classification models.

Muhammad Junaid Khan, Shah Hassan, Gita Sukthankar

Inspired by the recent success of transformers in natural language processing and computer vision applications, we introduce a transformer-based neural architecture for two key StarCraft II (SC2) macromanagement tasks: global state and build order prediction. Unlike recurrent neural networks which suffer from a recency bias, transformers are able to capture patterns across very long time horizons, making them well suited for full game analysis. Our model utilizes the MSC (Macromanagement in StarCraft II) dataset and improves on the top performing gated recurrent unit (GRU) architecture in predicting global state and build order as measured by mean accuracy over multiple time horizons. We present ablation studies on our proposed architecture that support our design decisions. One key advantage of transformers is their ability to generalize well, and we demonstrate that our model achieves an even better accuracy when used in a transfer learning setting in which models trained on games with one racial matchup (e.g., Terran vs. Protoss) are transferred to a different one. We believe that transformers' ability to model long games, potential for parallelization, and generalization performance make them an excellent choice for StarCraft agents.

Samaneh Saadat, Natalia Colmenares, Gita Sukthankar

The ever-increasing complexity of modern software engineering projects makes the usage of automated assistants imperative. Bots can be used to complete repetitive tasks during development and testing, as well as promoting communication between team members through issue reporting and documentation. Although the ultimate aim of these automated assistants is to speed taskwork completion, their inclusion into GitHub repositories may affect teamwork as well. This paper studies the question of how bots modify the team workflow. We examined the event sequences of repositories with bots and without bots using a contrast motif discovery method to detect subsequences that are more prevalent in one set of event sequences vs. the other. Our study reveals that teams with bots are more likely to intersperse comments throughout their coding activities, while not actually being more prolific commenters.

Ayesha Enayet, Gita Sukthankar

Good communication is indubitably the foundation of effective teamwork. Over time teams develop their own communication styles and often exhibit entrainment, a conversational phenomena in which humans synchronize their linguistic choices. This paper examines the problem of predicting team performance from embeddings learned from multiparty dialogues such that teams with similar conflict scores lie close to one another in vector space. Embeddings were extracted from three types of features: 1) dialogue acts 2) sentiment polarity 3) syntactic entrainment. Although all of these features can be used to effectively predict team performance, their utility varies by the teamwork phase. We separate the dialogues of players playing a cooperative game into stages: 1) early (knowledge building) 2) middle (problem-solving) and 3) late (culmination). Unlike syntactic entrainment, both dialogue act and sentiment embeddings are effective for classifying team performance, even during the initial phase. This finding has potential ramifications for the development of conversational agents that facilitate teaming.

Zerong Xi, Gita Sukthankar

This paper introduces a method for constructing an upper bound for exploration policy using either the weighted variance of return sequences or the weighted temporal difference (TD) error. We demonstrate that the variance of the return sequence for a specific state-action pair is an important information source that can be leveraged to guide exploration in reinforcement learning. The intuition is that fluctuation in the return sequence indicates greater uncertainty in the near future returns. This divergence occurs because of the cyclic nature of value-based reinforcement learning; the evolving value function begets policy improvements which in turn modify the value function. Although both variance and TD errors capture different aspects of this uncertainty, our analysis shows that both can be valuable to guide exploration. We propose a two-stream network architecture to estimate weighted variance/TD errors within DQN agents for our exploration method and show that it outperforms the baseline on a wide range of Atari games.

Abduljaleel Al-Rubaye, Gita Sukthankar

Popularity and engagement are the currencies of social media platforms, serving as powerful reinforcement mechanisms to keep users online. Social coding platforms such as GitHub serve a dual purpose: they are practical tools that facilitate asynchronous, distributed collaborations between software developers while also supporting passive social media style interactions. There are several mechanisms for "liking" content on GitHub: 1) forking repositories to copy their content 2) watching repositories to be notified of updates and 3) starring to express approval. This paper presents a study of popularity in GitHub and examines the relationship between these three quantitative measures of popularity. We introduce a weight-based popularity score (WTPS) that is extracted from the history line of other popularity indicators.

Samaneh Saadat, Olivia B. Newton, Gita Sukthankar et al.

Our goal is to understand the characteristics of high-performing teams on GitHub. Towards this end, we collect data from software repositories and evaluate teams by examining differences in productivity. Our study focuses on the team formation phase, the first six months after repository creation. To better understand team activity, we clustered repositories based on the proportion of their work activities and discovered three work styles in teams: toilers, communicators, and collaborators. Based on our results, we contend that early activities in software development repositories on GitHub establish coordination processes that enable effective collaborations over time.

Samaneh Saadat, Gita Sukthankar

While there are many machine learning methods to classify and cluster sequences, they fail to explain what are the differences in groups of sequences that make them distinguishable. Although in some cases having a black box model is sufficient, there is a need for increased explainability in research areas focused on human behaviors. For example, psychologists are less interested in having a model that predicts human behavior with high accuracy and more concerned with identifying differences between actions that lead to divergent human behavior. This paper presents techniques for understanding differences between classes of discrete sequences. Approaches introduced in this paper can be utilized to interpret black box machine learning models on sequences. The first approach compares k-gram representations of sequences using the silhouette score. The second method characterizes differences by analyzing the distance matrix of subsequences. As a case study, we trained black box supervised learning methods to classify sequences of GitHub teams and then utilized our sequence analysis techniques to measure and characterize differences between event sequences of teams with bots and teams without bots. In our second case study, we classified Minecraft event sequences to infer their high-level actions and analyzed differences between low-level event sequences of actions.

Saif Alabachi, Gita Sukthankar, Rahul Sukthankar

A physical selfie stick extends the user's reach, enabling the acquisition of personal photos that include more of the background scene. Similarly, a quadcopter can capture photos from vantage points unattainable by the user; but teleoperating a quadcopter to good viewpoints is a difficult task. This paper presents a natural interface for quadcopter photography, the SelfieDroneStick that allows the user to guide the quadcopter to the optimal vantage point based on the phone's sensors. Users specify the composition of their desired long-range selfies using their smartphone, and the quadcopter autonomously flies to a sequence of vantage points from where the desired shots can be taken. The robot controller is trained from a combination of real-world images and simulated flight data. This paper describes two key innovations required to deploy deep reinforcement learning models on a real robot: 1) an abstract state representation for transferring learning from simulation to the hardware platform, and 2) reward shaping and staging paradigms for training the controller. Both of these improvements were found to be essential in learning a robot controller from simulation that transfers successfully to the real robot.

Saif Alabachi, Gita Sukthankar, Rahul Sukthankar

Object detection models based on convolutional neural networks (CNNs) demonstrate impressive performance when trained on large-scale labeled datasets. While a generic object detector trained on such a dataset performs adequately in applications where the input data is similar to user photographs, the detector performs poorly on small objects, particularly ones with limited training data or imaged from uncommon viewpoints. Also, a specific room will have many objects that are missed by standard object detectors, frustrating a robot that continually operates in the same indoor environment. This paper describes a system for rapidly creating customized object detectors. Data is collected from a quadcopter that is teleoperated with an interactive interface. Once an object is selected, the quadcopter autonomously photographs the object from multiple viewpoints to %create training data that is used by DUNet (Dense Upscaled Net), collect data to train DUNet (Dense Upscaled Network), our proposed model for learning customized object detectors from scratch given limited data. Our experiments compare the performance of learning models from scratch with DUNet vs.\ fine tuning existing state of the art object detectors, both on our indoor robotics domain and on standard datasets.

Saif Alabachi, Gita Sukthankar

Drones are a versatile platform for both amateur and professional photographers, enabling them to capture photos that are impossible to shoot with ground-based cameras. However, when guided by inexperienced pilots, they have a high incidence of collisions, crashes, and poorly framed photographs. This paper presents an intelligent user interface for photographing objects that is robust against navigation errors and reliably collects high quality photographs. By retaining the human in the loop, our system is faster and more selective than purely autonomous UAVs that employ simple coverage algorithms. The intelligent user interface operates in multiple modes, allowing the user to either directly control the quadcopter or fly in a semi-autonomous mode around a target object in the environment. To evaluate the interface, users completed a data set collection task in which they were asked to photograph objects from multiple views. Our sketchbased control paradigm facilitated task completion, reduced crashes, and was favorably reviewed by the participants.

Alireza Hajibagheri, Gita Sukthankar, Kiran Lakkaraju

The aim of link prediction is to forecast connections that are most likely to occur in the future, based on examples of previously observed links. A key insight is that it is useful to explicitly model network dynamics, how frequently links are created or destroyed when doing link prediction. In this paper, we introduce a new supervised link prediction framework, RPM (Rate Prediction Model). In addition to network similarity measures, RPM uses the predicted rate of link modifications, modeled using time series data; it is implemented in Spark-ML and trained with the original link distribution, rather than a small balanced subset. We compare the use of this network dynamics model to directly creating time series of network similarity measures. Our experiments show that RPM, which leverages predicted rates, outperforms the use of network similarity measures, either individually or within a time series.

Liyue Zhao, Yu Zhang, Gita Sukthankar

Crowdsourcing platforms offer a practical solution to the problem of affordably annotating large datasets for training supervised classifiers. Unfortunately, poor worker performance frequently threatens to compromise annotation reliability, and requesting multiple labels for every instance can lead to large cost increases without guaranteeing good results. Minimizing the required training samples using an active learning selection procedure reduces the labeling requirement but can jeopardize classifier training by focusing on erroneous annotations. This paper presents an active learning approach in which worker performance, task difficulty, and annotation reliability are jointly estimated and used to compute the risk function guiding the sample selection procedure. We demonstrate that the proposed approach, which employs active learning with Bayesian networks, significantly improves training accuracy and correctly ranks the expertise of unknown labelers in the presence of annotation noise.